Intelligent Document Processing (Gartner)

Gartner published the Magic Quadrant for Intelligent Document Processing on 8 September 2026, only the second edition of the report, with vendor positions recorded as of August. Five vendors reached the Leader rung: ABBYY, Hyperscience, Infrrd, Tungsten Automation and UiPath.

The report is credited to Shubhangi Vashisth, Stephen Emmott, Tushar Srivastava, Sachin Joshi and Irina Guseva, and its abstract describes a field of fifteen vendors. Three of the five Leaders announced their own placements, and a fourth is named in published coverage of the results.

The 2026 Magic Quadrant for Intelligent Document Processing, and the five vendors on its Leader rung

UiPath is a Leader for the second consecutive year, and its account of the placement centers on a product that uses generative AI to turn complex unstructured documents into structured, verified data for what the company frames as agentic automation. ABBYY is a Leader for the second time, and its description of the work is the most specific of the five: a hybrid approach that pairs deterministic extraction with reasoning models, a capability it calls bring your own model, and a founding role in a document format specification. Tungsten Automation has now been a Leader in both editions of this quadrant, and its own numbers include more than ten billion documents processed annually across cloud, hybrid and on-premises deployments.

The second edition of a report is worth reading differently from a first. A first edition declares that a market exists. A second edition tells you who stayed, and the fact that Tungsten has held its position across both is a small piece of evidence about how slowly this market's top moves.

Automation Anywhere and OpenText were evaluated and are not on the Leader rung. Both are named in coverage of the results, and neither placement beyond that is established by the sources consulted, which is worth stating rather than filling in.

Hyperscience took both axes, and Forrester had already made it a Customer Favorite

Hyperscience's own announcement is the most specific claim in this quadrant: positioned furthest for completeness of vision and highest for ability to execute among all fifteen vendors evaluated, and first in three of the four use cases in the companion Critical Capabilities report, covering transactional document processing workflows, contextual intelligence workflows and specialized document processing workflows.

That is the same position Constructor took in the search and product discovery quadrant published eleven weeks earlier, and it is the shape a vendor occupies when it has no legacy portfolio pulling on its roadmap. Hyperscience sells document processing and not much else, which is why it can be first on vision among a field that includes ABBYY, OpenText and Automation Anywhere.

The cross-analyst counterpart is The Forrester Wave: Document Mining And Analytics Platforms, Q2 2026, published in May 2026 and authored by Boris Evelson, which scored eight providers against 31 criteria. UiPath and Hyperscience are the two Leaders confirmed by published material, and Forrester named Hyperscience the report's Customer Favorite. Iron Mountain is a confirmed Strong Performer.

So the two firms agree on the top of this market and disagree about how big it is.

Two firms drew the boundary at eight vendors and fifteen

Forrester scored eight providers in May 2026. Gartner scored fifteen in September. Neither number is wrong, and the gap between them is the most useful thing on this page for a buyer trying to work out who is even in the market.

The two firms also moved in opposite directions. Forrester's Q2 2024 edition scored fourteen providers against 25 criteria and named six Leaders, including UiPath, OpenText, expert.ai and Hyperscience, with Rossum as a Strong Performer. Its 2026 edition halved the field to eight. Gartner's second edition covers fifteen. One firm is narrowing this market toward its core and the other is widening it, and a shortlist assembled from only one of them will be missing vendors for reasons that have nothing to do with the buyer's requirements.

Part of the explanation sits in the technology. Document processing now overlaps with RPA, with agentic orchestration and with the automation platforms that own the workflow on either side of the extraction step, and a firm drawing its boundary conservatively will count fewer vendors as pure plays.

The category has two names and they share one word

Gartner calls this market intelligent document processing. Forrester calls it document mining and analytics platforms, a name that has survived four iterations on this site's own page covering it, through document-oriented text analytics and AI-based text analytics before that.

The shared word is document. Nothing else in the two names overlaps, and the two definitions do not line up either. Forrester defines its market as software that applies document and text mining and analytics technology to extract information from semi-structured documents that may include forms or document sections, which is a capability description. Gartner's name describes the outcome rather than the technique.

That difference matters when a buyer searches for a vendor list. The two documents will not find each other, and a procurement team that knows the market by one name will not see the other firm's Leaders at all.

Extraction keeps becoming a feature of the platform around it

The clearest structural fact in this quadrant is that its Leader rung overlaps with a different market's. UiPath is on it, and UiPath is also the name buyers know from robotic process automation. Automation Anywhere, another RPA vendor, was evaluated here and is not on the rung. A buyer already running an automation platform therefore has an extraction capability inside a product they own, and the question of whether to buy a document platform separately only arises when the documents are difficult enough that the built-in model loses accuracy.

That is why the vendors' own language has moved. UiPath describes its document product as feeding agentic automation. Tungsten Automation describes a unified platform for document automation, agentic orchestration and governance. ABBYY describes deterministic extraction paired with reasoning models. None of the three leads with extraction accuracy, and all three place the product inside a workflow that continues past the document.

For buyers, the practical consequence is that document processing is now a boundary decision before it is a vendor decision. A specialist will win on the hard tail of documents that a general platform cannot read, and will lose on the volume work that a platform the buyer already licenses handles adequately at no incremental licence cost. The quadrant cannot make that call, because both kinds of vendor sit on the same chart and the chart does not know what the buyer already owns.

What the use-case scores say that the quadrant does not

Hyperscience's first place in three of four use cases is a more specific claim than its Leader placement, and the two should not be read as the same statement. A Leader placement describes a vendor's position across the whole evaluation. A use-case score describes how a vendor performs when the buyer's problem is narrowed to one shape of work.

The four use cases in Gartner's companion report split document processing along lines a buyer will recognize. Transactional workflows are the high-volume, structured-by-convention documents where accuracy and throughput decide the business case. Contextual intelligence workflows are the ones where the document has to be understood rather than parsed. Specialized workflows are the long tail that no general platform handles well.

The distinction is what turns a quadrant into a shortlist. A vendor fourth on the aggregate can be first on the use case that matches the buyer's actual intake, and the reverse is equally true.

What a price per page does to a document project

Forrester's Q2 2026 edition is notable for stating a price, which this site's page on that report covers in detail: roughly five cents a page at volume, against accuracy that begins well below what a production deployment needs and improves with tuning. That is the first time either firm has put a number in the document, and it changes the shape of the conversation.

A unit price turns a document processing decision into a labor-cost comparison, because the alternative to a five cent page is a person, and the person's cost per page is knowable. It also makes the tuning period visible as a cost rather than a phase, since accuracy that starts low and improves means the first months of a deployment are the most expensive per correctly extracted document that the program will ever have.

Gartner's quadrant does not publish a price. The two documents are therefore complements rather than substitutes for a buyer building a business case, and a procurement team that reads only the placement will struggle to answer the question its finance function will ask first.

What to ask before you buy document processing in 2026

Ask which of the two names your own team uses. If the requirement is written as document mining and analytics, the Forrester Wave is the closer instrument, and if it is written as intelligent document processing, Gartner's quadrant is. The two will not surface each other's Leaders, and the buyer who knows only one name will not know the other list exists.

Ask for the use-case score alongside the quadrant position. Four use cases and a Critical Capabilities report produce a ranking that moves with the shape of the intake, and the aggregate placement conceals exactly the variation a buyer needs.

Ask what the platform does when the document is not a form. The tail of specialized documents is where general platforms lose accuracy, and it is where a buyer discovers the true cost of a pilot that was run on clean samples.

Ask whether the vendor is being scored as a document platform or as part of an automation suite. Three of the five Leaders here sell extraction into a wider automation or process platform, which is a different purchase from a specialist and a different integration cost.

Ask what the first year costs per correctly extracted page rather than what the licence costs per page. Those two numbers separate once tuning time is priced in, and the gap between them is the honest answer to whether a program pays for itself.

Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner's research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.

The market this one keeps growing into is covered separately. Robotic Process Automation is the automation category that owns the workflow on either side of the extraction step, and the vendor leading this quadrant is also a name buyers know from that one. The boundary between the two moves every year, and it moves in one direction: extraction becomes a feature of the platform rather than a product a buyer procures on its own.

The highest-volume application of this technology is a market in its own right on this site. Accounts Payable Invoice Automation Software (Forrester) covers the invoice intake workflow, which is where most document processing programs start and where the accuracy and cost-per-page figures in this article are actually proved. A buyer evaluating both should read them in that order, since the document platform is chosen to serve a process that the finance function already owns.